Review of roll-to-roll fabrication techniques for colloidal quantum dot solar cells

Energy is directly linked to the well-being and prosperity of all countries and is a determinant for the industrialization and quality of human life. However, traditional energy sources, like coal, oil, and natural gas, are non-renewable, and their unstable prices usually cause disturbance globally. In the meanwhile, the heavy dependence on non-renewable energy is causing climate change and environmental crisis, which has attracted universal attention. One solution to this issue is to develop renewable energy technologies, such as wind, solar, as well as geothermal and tidal energy based solutions. Amongst, solar energy [1] is especially attractive, because it possesses numerous advantages including clean, free, abundance, and more uniform distribution around the globe compared with other renewable energy and it can be accessed from the space as well [2].

Solar energy can be converted into various types of energy products including electricity [3-6], chemical fuel [7-13], and heat [14,15]. The conversion of solar energy into electricity can be accomplished by solar cells. Recently, the power conversion efficiency (PCE) of different types of solar cells is surging. For example, the lab-scale organic solar cells, copper indium gallium selenide (CIGS) solar cells, colloidal quantum dot (CQD) [16] solar cells, and perovskite solar cells have achieved the highest PCE of 19% [17], 23% [18,19], 18% (from the Best Research-Cell Efficiency Chart of National Renewable Energy Laboratory (NREL)), and 25% [20,21], respectively.

Among these solar cells, CQD solar cells have attracted broad attention. As shown in Fig. 1, for lab-scale CQD solar cells, continuous improvement can be observed and as high as 18% of PCE has been certified. It is also prospective to surpass the 20%-PCE milestone, which will then be great potential in commercial applications. One unique property of CQDs is that their bandgaps can be tuned by simply changing the crystal size, so that they can harvest solar radiation in a broad range including the near-infrared (NIR) region [22,23]. As shown in Fig. 2, lead sulfide (PbS) CQDs of different sizes can absorb radiation of different NIR windows, far beyond those of silicon (Si). Thus, CQD solar cells can theoretically realize higher efficiency than other types of solar cells or augment the existing solar cell technologies by harnessing the NIR solar spectrum with a tandem architecture [22,24-38].

Figure 1.  Progress of certified PCE of CQD solar cells (data are extracted from the NREL website: https://www.nrel.gov/pv/cell-efficiency.html).

Figure 2.  Comparison of the solar radiation absorption range between Si and PbS CQDs.

Currently, several CQD material systems have been under extensive investigations, where lead chalcogenide CQDs [39-41] and perovskite CQDs [42,43] are attracting more and more attention due to their better overall performance. In this review, we mainly focus on lead chalcogenide CQD solar cells, which are mostly based on PbS [40] and lead selenide (PbSe) [41] because both of them are capable of harnessing NIR radiation. The highest PCE values of ~14% [44] and 11% [45,46] have been reported in PbS and PbSe CQD solar cells, respectively, and >1.5% [23] and 1.2% [47] in Si-filtered PbS and PbSe CQD NIR solar cells, respectively. This means that additional 1% of PCE could be potentially added to the state-of-the-art Si solar cells.

The steady improvement of CQD solar cells is the synergistic result of materials science, device engineering, and theoretical studies. Pristine synthesized CQDs [48,49] are capped with insulating ligands for better dispersity in nonpolar solvents, which requires subsequent ligand exchange processes for shorter conductive ligands [41], such as halides [50], mercaptopropionic acid (MPA) [51], and 1,2-ethanedithiol (EDT) [52]. The earlier in-place solid phase ligand exchange of CQD thin films needs layer-by-layer (LbL) steps to deposit thick films with a few hundred nanometers, which is time consuming and wasting. With the development of the solution phase ligand exchange (SPLE) [50], CQD inks make it available to directly fabricate thick CQD thin films with a one-step process [53,54]. Recent advances in the CQD ink technology allow for improved passivation of CQD surfaces and facile one-step deposition of high-quality CQD thin films, thus contributing to highly efficient CQD solar cells and also paving the way to fabricate high-performance large-area CQD solar cells with more advanced roll-to-roll compatible coating or printing techniques.

This review focuses on the developments of the large-area fabrication of CQD solar cells. It mainly covers the progress of typical large-area coating and printing techniques, including spin coating, spray coating, blade coating, dip coating, and slot-die coating. First, earlier developments on CQD inks and other direct deposition techniques are introduced. Then for each deposition technique, the experiment setups, deposition protocols, and underlying mechanisms are outlined. After that, the corresponding progress of CQD solar cells based on each deposition method is discussed and summarized. Finally, future steps are presented along with new strategies to accomplish the ultimate goal of the low-cost large-area fabrication of CQD solar cells.

The fabrication methods of CQD thin films can be classified into two categories: LbL deposition [41] and one-step deposition [53]. The LbL method is usually fabricated with hot-injection synthesized pristine CQD solutions, while the one-step method employs CQD inks. In this section, we will detail the protocols of LbL and one-step methods, and also outline recent developments in CQD ink technologies.

The LbL method usually employs CQDs which are synthesized with the hot-injection method [48,55,56] (see Fig. 3) and capped with long insulating carboxylic ligands. Aiming to achieve conductive CQD thin films, it is necessary to remove those insulating ligands and replace them with more conductive ligands. This is generally termed the solid-state exchange (SSE), since this ligand exchange involves in-place solid-liquid reactions. A typical SSE procedure is shown in Fig. 4. First, a thin CQD film (~30 nm) capped with carboxylic ligands is formed with various coating and printing methods. Then a solution containing another ligand precursor is brought in contact with the film, where ligand exchange reactions happen. As the new ligands can only penetrate a limited depth, a thin initial layer is required. Subsequently washing steps are carried out several times to remove the remaining ligands leaving on the surface of the exchange film, which makes the exchanged CQD film usually thinner than the pristine film. To achieve a thick CQD film, it is necessary to repeat the above procedures multiple times.

Figure 3.  Schematic of hot-injection synthesis.

Figure 4.  Schematic of the SSE process.

The highest PCE of 12.44% was achieved by the LbL-based PbS CQD solar cell [57], which was prepared by the following LbL protocols. First, to prepare the precursor solution, an iodide source, 1-ethyl-3-methylimidazolium iodide (EMII), was dissolved in methanol with a concentration of 27 mM. Then the pristine PbS CQD thin film was formed on the substrate by spin coating the 100 mg/mL PbS colloidal solution at a spin speed of 2500 revolutions per minute (rpm). After that, the EMII precursor solution was dipped onto the top of the pristine film and left standing for 30 s, during which oleate ligands were replaced with iodine ligands. Finally, two rinse-spin steps with methanol and acetonide (ACN) were successively conducted. Thereafter a single deposition cycle was completed and a PbS CQD layer with the thickness of ~30 nm was thus formed. It is necessary to repeat this cycle multiple times to achieve the desired film thickness, such as, 10 cycles to achieve a 300 nm CQD layer.

Obviously, the LbL method requires at least 50 spin-coating cycles to form a thick CQD layer. Assume that one spin-coating cycle takes 30 s, and thus a CQD active layer with the thickness of 300 nm needs more than 1500 s or 25 min, even though the efficiency and quality of the ligand exchange are not considered. Moreover, cracking is easily generated as the LbL-deposited film becomes thicker and it is very challenging to make a crack-free large-area film [58]. Thus, innovative and time-saving ways are under exploration to realize high-quality CQD thin films.

To overcome the above limitations, CQD inks are proposed [46,50,53,54,59,60]. There are two feasible ways to make CQD inks: SPLE [28] and direct synthesis [46,59,60], where SPLE is adopted extensively. Because hot-injection based CQDs are used in SPLE, which enables that the size dispersity of the CQD ink can be largely preserved [54]. However, the cost of CQD inks synthesized by this SPLE route is averagely high to 35 USD/g. Although it can be reduced to nearly 5 USD/g by the direct synthesis method [59], the size tunability of the direct synthesized CQD ink remains to be improved. In this subsection, we will detailedly introduce the differences between those two methods of synthesizing CQD inks.

Fig. 5 shows a typical preparing process of SPLE.

Figure 5.  Schematic of SPLE process.

The precursor solution is prepared first, where N,N-dimethylformamide (DMF) is usually adopted as the solvent and lead halides (PbX) as the new ligands, and also sometimes additives are required [54]. For example, Sun et al. prepared an exchange precursor solution with a composition of 0.1 M PbI₂, 0.04 M PbBr₂, and 0.04 M ammonium acetate in DMF [44].

Then the precursor solution and the pristine CQD solution (usually in octane or hexane) are vigorously mixed for the ligand exchange reaction. As DMF and the nonpolar solvent (e.g., hexane or octane) are orthogonal to each other, after the ligand exchange, CQDs will move into the bottom DMF phase and can be easily separated by removing the upper nonpolar phase with the pipette. Subsequently, the bottom DMF phase is washed with the nonpolar solvent to remove the residual organic ligands. During this step, the same amount of the nonpolar solvent is added into the DMF phase, then the mixture is vigorously mixed, and, after the phase separation, the upper nonpolar phase is removed. Then the antisolvent is mixed into the DMF phase and the mixture is centrifuged to precipitate the CQD solids, which are dried and re-dispersed in the final solvent, such as butylamine (BTA), propylene carbonate (PPC), 2,6-diflfluoropyridine (DFP), or their mixtures (for example, BTA and DMF). To achieve better CQD films, other additives, such as the mixture of cesium iodide and lead bromide, are usually added to modify the ink. Notably, in order to realize a thick CQD film with this one-step coating process, a high-concentration ink is needed (such as, 300 mg/mL).

For example, in Sun et al.’s report [44], the tube containing the mixed precursor solution and the pristine CQD octane solution was vigorously shaken for 2 min to mix these two solutions, during which CQDs were completely transferred to the DMF phase, and then the mixture was allowed to stand for 30 s to separate the two phases. The upper octane phase was discarded and the DMF phase was then washed three times by vigorously mixing with 5 mL octane for 30 s, followed by adding 3 mL toluene and centrifuging at 3300 rpm for 1 min to precipitate CQDs. The solids were then dried for 15 min in the vacuum antechamber of a glovebox and redispersed in BTA to get the final CQD ink.

Finally, the CQD film is obtained by spin coating with the ink. The spin speed ranging from 1500 rpm to 2500 rpm is adopted to produce CQD films with different thicknesses in air. Usually multiple spinning steps should be executed, such as a low-speed initial spinning step to spread the ink, then a moderate-speed spinning step to achieve the desired thickness, and finally a high-speed spinning step to remove the residual solvent. This usually produces a highly smooth and reflective CQD film. However, in a certain circumstance, an additional annealing step is needed. Typically with a 15 min annealing treatment at 70 °C in a N₂ filled glovebox, a high-quality PbS CQD film is made.

Compared with the LbL method, depositing CQD films with CQD inks by SPLE usually takes much less time. In a typical fabrication process, it only takes a few minutes to prepare a CQD film with the thickness of 300 nm. This is a huge advance in terms of manufacturing simplicity and time consumption.

For the direct synthesis of CQD inks as shown in Fig. 6 [59], lead (Pb) and sulphur (S) precursors, such as lead iodide (PbI₂) and thiourea, are dissolved in DMF. Then an amine, such as BTA, is mixed into the precursor solution and the solution quickly turns from yellow to dark black, yielding the halide-passivated CQD ink in seconds. Since the synthesis is performed at room temperature (RT), this ink preparation method is also termed the room temperature one-step (RTOS) synthesis.

Figure 6.  Schematic of direct synthesis of the CQD ink.

Both SPLE and direct synthesis methods have the capability to produce CQD inks with a high concentration of >200 mg/mL, rendering a thick CQD film with one spin-coating step. Thus its fabrication time can be reduced and only a few minutes are required. This is more advantageous than LbL and has been widely used in the CQD solar cell community. Especially for the direct synthesis method, not only is it quick and easy, but it also can realize high-performance CQD solar cells. For example, Ma’s group demonstrated over 10% PCE for both PbS [59,61] and PbSe [46] CQD solar cells based on direct-synthesized CQD inks. Those PCE values are comparable to that of the state-of-the-art SPLE-based ones, making the direct synthesis method highly attractive for both academia and industry. This kind of CQD ink technology also enables it viable to fabricate high-quality large-area CQD films and solar cells, which will be discussed in detail in the following sections.

There are various coating methods for liquid films, which have been extensively used in industry fields, however, only a few of them are capable of producing high-quality semiconductor thin films. In this section, we will review these coating methods which are frequently used to produce semiconductor films from precursor solutions or inks, including spin coating, spray coating, blade coating, dip coating, and slot-die coating. For CQD solar cells, spin coating is the most widely used coating method, followed by spray coating and then blade coating. However, regarding to slot-die coating, there have been no reports, thus we will introduce the principles and examples from other solar cell technologies to exhibit its potential in large-area CQD solar cells. Table 1 summarizes the material and device performance of major CQD solar cells reviewed in this work [44,57,61-69], where V_oc is the open-circuit voltage, J_sc is the short-circuit current density, and FF denotes the fill factor.

Spin coating is the first choice for CQD solar cell research. The best-performed CQD solar cells [42,44] are fabricated with this method and a few important concepts [41,52,70] are being developed. Though the film formation mechanism of spin coating is different from that of other large-area deposition methods, it provides a reliable platform for exploring materials science and device physics. This subsection mainly discusses the underlying principles of spin coating and emphasizes its difference from other deposition techniques.

Spin coating is a fluid flow process [71-76] and is governed by the combined effects of convection and diffusion. As shown in Fig. 7, during the spin-coating process, inks or solutions are dispensed on a substrate which is mounted on a rotating disk. As the disk starts to accelerate, the inks or solutions spread, the solvents evaporate, and thus the solutes or suspensions are left, finally forming a uniform film. Several theoretical studies on spin coating of colloidal suspensions have been reported [73-75]. The effects of the angular velocity, initial solvent weight fraction, solvent property, and spin-coating protocol on the evolution of temperature and concentration profiles in the liquid film during the spin-coating process have been examined numerically and verified experimentally. In general, the final film thickness (h) is found to vary with the angular velocity (Ω) according to $h = {\Omega ^{ - 1/2}}$ [73], and also depends on the solution viscosity and diffusivity. In addition, by controlling the substrate temperature, solvent mixture, or surrounding atmosphere, it is possible to further adjust the film properties, especially the film morphology, which is vital for high-quality solar cells. Thus, optimizing the ink or solution properties and spin-coating protocols, including the spinning speed, acceleration, spinning steps, and heating, is a potential avenue to achieve a good film.

Figure 7.  Schematic of the spin-coating setup and process.

For example, the final film thickness could be affected by the solvent remarkably. With a low boiling point solvent, such as BTA, a PbS CQD ink with a concentration of 200 mg/mL led to a PbS CQD film with the thickness of ~350 nm at a spin-coating speed of 2500 rpm [54]. While with a high boiling point solvent, such as DMF, the film thickness was ~250 nm at a lower spin-coating speed of 2000 rpm, even though under the same CQD concentration [77]. It indicates that a lower boiling point solvent results in a thicker film. As for the same solvent, the thickness could be effectively tuned by changing the spin speed or concentration. In Ref. [77], Ma’s group further investigated how the concentration of a DMF solvent based PbS CQD ink affects the CQD film thickness. As shown in Fig. 8, the thickness varies approximately linearly with the ink concentration. This is similar to that found in other types of solution-processed thin films, such as the ZnO sol-gel [78] or polymer [79].

Figure 8.  Concentration-thickness dependence of the PbS CQD ink, ZnO sol-gel, and P3HT:IDTIDT-IC polymer solution.

The highest PCE value in PbS CQD solar cells was recently reported by Sun et al., which was fabricated by spin coating with the SPLE-based PbS CQD ink [44]. In this work, the authors proposed a post-treatment method (see Fig. 9) to improve the electrical transport of the CQD film, and demonstrated that the post-treatment of the spin-coated CQD film with a perovskite-type precursor solution could induce surface chemistry that converts the less conductive materials (also named as matrix) surrounding CQDs into a high conductive matrix, thus improving the carrier transport and leading to better device performance.

Figure 9.  Strategy and performance of spin-coated PbS CQD solar cells by Sun et al. Reproduced with permission [44].

Although various deposition methods based on spin coating have been successfully developed, most of them were carried out on small lab-scale substrates and difficult to obtain large-area films with good uniformity and a sufficient thickness. As an alternative method, spray coating could precisely control the thickness and uniformity of CQDs films and is an ideal way for mass production.

Fig. 10 shows a typical spray-coating process where small droplets dispersed in the carrier gas are collected by the substrate [80]. It is a multi-step process, including the atomization of liquid solutions or mixtures, droplet flight and evaporation, droplet impact on the substrate, droplet diffusion, retreat, recoil, drying, solute adhesion, and self-bonding with the substrate [81]. To obtain high-quality sprayed films, all of these steps must be well understood and precisely controlled.

Figure 10.  Schematic of spray coating.

In the spray-coating process, the liquid precursor is atomized and sprayed by the pressurized gas in the spray gun, the tiny liquids drop onto the substrate (usually being heated), the solvent evaporates rapidly, and thus the solute deposits on the substrate to form a thin film. The whole process can be done under atmospheric environmental conditions. Uniform thin films can be achieved by simply controlling gas pressure via injecting the high-pressure gas into the nozzle, and the thickness is controllable and determined by spraying cycles. Compared with spin coating, spray coating is a very effective and expandable method for depositing thin films on both hard and flexible substrates. Besides, it also exhibits several advantages in the field of CQD solar cells: First, the utilization rate of precursor solutions or inks in the spraying process is ~100%, contributing to the reduction in the production cost; second, spray coating is applicable to almost all surfaces, making it suitable for flexible substrates based CQD solar cells, wearable devices, automobile surfaces, or urban infrastructure. These enable spray coating widely used in various CQD solar cell studies.

In 2010, Im et al. demonstrated a CQD sensitized solar cell with spray-coated CdSe CQDs having a ZnS glue layer [82]. They showed that with spray coating, CdSe CQDs can penetrate into the bottom most TiO₂ layer, achieving uniform coverage of mesoporous TiO₂ (mp-TiO₂) films. With annealing at 450 °C for 5 min, CQDs were in close contact with the mp-TiO₂ layer, which is advantageous to the photocarrier transfer process within the solar cell. As a result, this CQD sensitized solar cell realized V_oc of 0.57 V, J_sc of 11.2 mA/cm², and FF of 0.35, corresponding to PCE of 2.2%. Although they did not investigate the impact of spraying parameters on the film properties, this study demonstrated that spray coating is feasible to prepare uniform coatings and fill the pores of porous materials.

In 2015, Kramer et al. demonstrated a PbS CQD solar cell fabricated by spray coating for the first time [63]. They used oleate-capped PbS in octane as the CQD solution. With low-concentration PbS CQDs (3.33 mg/mL), they showed that a monolayer-thick PbS CQD film was precisely formed. By developing an automated spraying-based LbL setup, they fabricated PbS CQD solar cells with a depletion-heterojunction structure [70] and achieved average PCE of 6.5% with champion PCE of 8.1%, comparable to the average PCE value (6.7%) of spin-coated devices (see Fig. 11). Further, they also fabricated PbS CQD solar cells on flexible substrates, such as polyethylene terephthalate (PET), and achieved PCE of 7.2% [83]. These studies indicate that spray coating has great potential to make solar cells on arbitrary substrates.

Figure 11.  Current density versus voltage curves (left) and V_oc, J_sc, and FF values as a function of operating time (right) of different batches of spray-coated PbS CQD solar cells from Ref. [83]. Reproduced with permission [83].

Fig. 12 highlights the key features of the spraying setup [83]. There are four spraying nozzles, with one functioned as the CQD source, and the others used for SSE, rinsing/washing, and drying, respectively. All spraying nozzles could be opened or closed to allow that the step-by-step SSE process could be repeated in a precise time sequence. In order to achieve a CQD active layer with full surface coverage, 65 cycles–85 cycles are required. In this setup, the solvents used are the same as those used for spin coating, while the differences are the use of low-concentration CQD for deposition and dry air to remove the residual solvent from the film. All these indicate spray coating is a versatile method where the recipes of spin-coating methods could be directly employed.

Figure 12.  LbL spraying setup for CQD solar cells from Ref. [83]. Reproduced with permission [83].

Later in 2016, Park et al. also employed spray coating to fabricate PbS CQD solar cells but with a different coating recipe [64]. Instead of multiple LbL cycles, a one-step SSE method was applied. As shown in Fig. 13, the PbS CQD octane solution with a much higher concentration of 50 mg/mL was adopted, and thus a thick PbS layer was sprayed. The sprayed film was dipped into a precursor solution for SSE, where the solution was composed of mixed tetrabutylammonium iodide (TBAI) and EDT ligands with a specific TBAI/EDT ratio of 15:4. With an optimized SSE time period, the prepared solar cells achieved PCE of 4%, V_oc of 0.57 V, J_sc of 11.79 mA/cm², and FF of 0.60. The current density versus voltage curves for the devices with different ligand exchange time periods are shown in Fig. 14. It shows that with the exchange time of 50 s, the device performed best, indicating that a moderate interval of soaking could achieve a good ligand exchange result. Although due to its thinner active layer of 140 nm, inferior performance with much lower J_sc was resulted, compared with the devices (>200 nm) made by Kramer et al. [63], the fabrication time of the CQD layer was greatly reduced. Because only one ligand exchange cycle was conducted in this method while more than 10 exchange cycles are required in the common LbL process. It can be expected that the performance will be improved by further optimizing the device structure and layer thickness.

Figure 13.  Spray-coating protocols for PbS CQD thin films by Park et al. Reproduced with permission [64].

Figure 14.  Current density versus voltage curves of spray-coated PbS CQD solar cells based on one-step SSE from Ref. [64]. Reproduced with permission [64].

As the fast development of the CQD ink technology, different avenues to produce high-quality CQD solar cells with spray coating also emerge. With CQD inks, it is not necessary to conduct the ligand exchange or solvent rinsing/washing on the sprayed CQD film. That is to say, it is possible to fabricate CQD solar cells with only a single spraying step, leading to a significant reduction in both time and cost.

Choi et al. first demonstrated the viability of CQD inks for spray coating CQD thin films [84]. They adopted SPLE to make the CQD ink used for the spraying process, during which a perovskite-type ligand composition was used to replace the long oleate ligands, and found that this CQD ink could be colloidally stable under ambient conditions for months. This is because an electrical double layer comprising PbI₃ anions and MA cations was formed around CQDs. Fig. 15 shows the corresponding setup with supersonic spraying employed. During the spraying process, the air from the supersonic nozzle could be heated to the desired temperature, allowing rapid evaporation of the solvent and facilitating the deposition of the PbS QD ink, which eliminates the necessity of the annealing treatment after deposition. Heat and pressure were converted into kinetic and adhesion energy of the CQD droplets, allowing for the direct deposition of PbS QDs without using any binder material. As a result, only the dried and crystallized CQD particles hit the substrate and formed a dense and smooth film (average film roughness of 3.4 nm) with superior adhesion. As shown in Fig. 16, this spray-coated PbS CQD solar cell achieved a slightly lower PCE value of 3.7%. This is mainly due to its unoptimizable CQD ink and device structure.

Figure 15.  Schematic of spraying deposition of PbS QD inks by Choi et al. Reproduced with permission [84].

Figure 16.  Device performance of the PbS CQD solar cell from Ref. [84]: (a) current density versus voltage curve and (b) external quantum efficiency (EQE) curve. Reproduced with permission [84].

For spray-coating methods with CQD inks, the parameter control is especially critical. In 2019, Choi et al. demonstrated scalable fabrication of CQD films by tuning the solute-redistribution dynamics of CQD inks with an ultrasonic spray-coating system [65]. The uniform deposition of CQD inks was obtained by manipulating the substrate temperature. When a substrate temperature of 50 °C was applied, the evaporation of the solvent was expedited, reaching an evaporation rate of 2.67 µL/min which is 6.4 times faster than that at RT (0.42 µL/min). At an appropriate evaporation rate, the contact line receded quickly, and thus CQDs were deposited on the substrate rather than piled with the receding contact line, finally realizing the uniform deposition (as shown in Fig. 17 (a)). With such a spraying protocol, the authors successfully fabricated a hero device with an ITO/ZnO/CQD-MAI/CQD-EDT/metal electrode structure. Here ITO is indium tin oxide, MAI is methylammonium iodide, and the thicknesses of the ITO, ZnO, CQD-MAI, CQD-EDT, and metal electrode layers are 180 nm, 100 nm, 265 nm, 65 nm, and 160 nm, respectively. As shown in Figs. 17 (b) and (c), this device could achieve J_sc of 24.1 mA/cm², V_oc of 0.56 V, FF of 0.6, and PCE of 8.10% under AM1.5G conditions.

Figure 17.  Device fabrication process and performance of PbS CQD solar cells from Ref. [65]: (a) images of devices at different fabrication stages, (b) PCE statistics, and (c) current density versus voltage curve and SEM image indicating the device structure. Reproduced with permission [65].

Another advantage of spray-coating methods with CQD inks lies in more solvent choices used for the CQD ink. The most widely used and best-performed CQD inks were formed with volatile BTA, and able to realize high-performance PbS CQD solar cells with PCE of larger than 13.8% by the spin-coating method [44]. However, BTA is incompatible with scalable deposition methods, such as spray coating. Because its high vapor pressure usually makes solvent evaporation uncontrollable, resulting in an uneven film. To overcome this difficulty, Yang et al. proposed a mixed solvent system [66]. By mixing BTA with DMF (Fig. 18 (a)), they showed that the binary solvent system has good colloidal stability and appropriate vapor pressure, availing to form stable CQD inks. Because the medium acidity and alkalinity of DMF in the mixed solvent could effectively prevent the CQD surface from destroying by alkali (base) and improve the colloidal stability of CQD inks by maintaining the electrostatic charge balance and preventing aggregation. It demonstrated that the mixed solvent system exhibited excellent processability for spray coating, and could repeatedly manufacture high-quality CQD films with minimal macro-defects (Fig. 18 (b)). As shown in Figs. 18 (c)−(f), the corresponding PbS CQD solar cell fabricated by this spray-coating method exhibited better performance with resulted PCE of 8.84%. This is also the highest PCE level that has been reported in spray-coated PbS CQD solar cells. This research can be served as a valuable reference to design new solvent systems for CQD inks.

Figure 18.  Spray-coating process and device performance of PbS CQD solar cells from Ref. [66]: (a) schematic of spray-coating process; (b) images of solar cell devices at different fabrication stages; (c) current density versus voltage curves, (d) PCE statistics, (e) transient photovoltage comparison, and (f) transient photocurrent comparison of solar cells made from different solvents. Reproduced with permission [66].

Blade coating is another widely used large-area thin film fabrication method for CQD solar cells, which is fast, simple, and straight forward. Table 2 compares the time consumption for the fabrication of a 1-inch-by-1-inch film with CQD inks. Obviously, blade coating offers the fastest manufacturing speed, which only takes ~1 s to make such a high-quality film. Fig. 19 shows a schematic of blade coating, in which the blade is usually mounted on a moveable track. As the blade travels along the plane parallel to the substrate, the CQD ink is extended over the substrate. The thickness of the CQD film can be controlled easily by adjusting the distance between the blade edge and the substrate surface, and the film properties could be tuned by other processing parameters, such as substrate heating and the surrounding environment. Since the substrate is usually fixed, an active drying process might be necessary in the blade-coating method. However, blade coating is still superior to spin coating in terms of the consumption of both time and cost. For example, the utilization rate of CQD inks in blade-coating methods is extremely high, and usually, only a small amount of 5% is wasted. Here we will further introduce recent studies focused on CQD solar cells prepared by blade coating.

Figure 19.  Schematic of blade coating.

Blade coating has the advantage of easily forming thick CQD active layers. Compared with spin coating, the thickness of the blade-coated film can be controlled easily by tuning the blade edge height and the coating speed. The additional control makes it easier to achieve a thicker film with the same solution concentration, allowing for more device engineering methods. In 2020, Fan et al. demonstrated an advanced blade-coated PbS CQD solar cell by combining new ligand exchange, solvent engineering, and the blade-coating strategy [23]. With blade coating, smooth micron-thick CQD active layers were achieved, which are very difficult for spin coating. However, this is not enough to get high-performance solar cells, as thicker active layers require better exciton diffusion, which is exactly one of the major bottlenecks to achieve higher efficiency. However, the SPLE protocol developed for smaller CQDs was found to be not applicable for larger CQDs. Because with traditional SPLE, the dispersivity of larger CQDs is poor. Fan et al. developed new SPLE methods by introducing an additional resolvation step after the initial SPLE to improve the dispersivity. Furthermore, they found the widely used solvent for redispersing CQDs would cause the film cracked, once the thickness exceeds 1 μm. Thus they also updated the solvent system with a mixture of DMF, dimethyl sulfoxide (DMSO), BTA, and phenethylamine. With all these three innovations, they achieved the thickest (1.2 μm) PbS CQD solar cell which has been reported. This solar cell is capable of reaching EQE of 80% with both the first exciton peak and the nearest Fabry-Perot resonance peak in the infrared region between 1100 nm and 1800 nm, which is a feat in CQD solar cells.

Obviously, solvent engineering is of equal importance for blade-coated high-quality CQD films. As we have mentioned above, during the blade-coating process, the CQD ink is first deposited on the blade edge to form a meniscus between the substrate and the blade edge (as shown in Fig. 19). Then the blade is pulled over the substrate. The blade should move at a high speed (>25 mm/s) in order to enable the film to be formed in the Landau-Levich regime [85,86]. Since the coated CQD film is wet, additional heating was needed to remove the excess solvent in case island-like spots formed in the films. The film undergoes an initial solidification process when heating, which can prevent the ink from further migrating on the substrate. It indicates that the solvent plays a key role in the drying process. Take DMF/BTA with a ratio of 30/70 as an example. This solvent can be dried very quickly, however, cracked films will form with a high concentration of CQDs (>250 mg/mL). With a lower concentration, the film will be dried nonuniformly, resulting in rough, macroscopic textures across the top of the film. With the proposed quaternary solvent composed of DMF, DMSO, phenethylamine, and BTA, Fan et al. demonstrated that adding solvents with a higher boiling point to slow down the evaporation of the wet film is a feasible way to realize a smooth film. Based on this, they adopted an ITO/ZnO/PbX-PbS/EDT-PbS/Au architecture and constructed an IR harvesting device. As shown in Fig. 20, this device could realize J_sc of 38.9 mA/cm², accounting for ~60% of total solar photons under AM1.5G illumination. This is the highest level that has been reported in solution-processed thin-film solar cells. Moreover, the corresponding Si-filtered J_sc (IR-J_sc) reached 9.81 mA/cm², 70% larger than that achieved by previously reported IR-CQD solar cells, which is mainly due to the very high EQE in the IR region (Fig. 20 (d)). This also enables the best-performed Si-filtered CQD solar cell with PCE of 1.57% [23].

Figure 20.  Device structure and performance of the micron-thick PbS CQD solar cell by Fan et al.: (a) SEM image; current density versus voltage curves under (b) AM1.5G solar radiation and (c) Si-filtered AM1.5G solar radiation; (d) EQE curve of the PbS CQD solar cell. Reproduced with permission [23].

Similarly, in 2021, Sukharevska et al. investigated the solvent effect for blade coating with the PbS CQD ink [67]. Generally, the dispersion of the PbS CQD ink (with BTA as the solvent) suffers from colloidal instability. In this work, Sukharevska et al. found 2,6-difluoropyridine (DFP) could well disperse PbS CQDs capped with CH₃NH₃PbI₃ ligands and the inks could be colloidally stable for more than 3 months. DFP could also yield stable dispersion for large-diameter PbS QDs, thus extending the coverage to the NIR region. By optimizing the blade-coating parameters (the blade height and speed) of DFP-based inks, high-quality films were achieved when the substrate was heated at low temperatures (70 °C−100 °C). As a result, PbS CQD solar cells with PCE high to 8.7% were demonstrated, as exhibited in Fig. 21. This work also indicates that blade coating can be applicable for the active layer deposition with no sacrifice of device efficiency.

Figure 21.  Performance of solar cells in Ref. [67]: (a) current density versus voltage curves and (b) EQE and integrated J_sc curves. Reproduced with permission [67].

Later, the same group fabricated efficient p-n junction PbS CQD solar cells with a structure of ITO/MPA-PbS/PbX-PbS/ZnO/Al, where both the p-type and n-type PbS CQD thin films were prepared by blade coating with respective inks [68]. The p-type ink used in this work was composed of MPA-capped PbS CQDs in a mixed BTA:water solvent. With an optimized BTA:water ratio of 50:50, the ink could be stable for several hours for further processing. This ratio could also enable a uniform coating of the p-type ink with the overall roughness of 6.1 nm. They found that with a higher water ratio, the wettability of the ink was unapplicable to blade coating, while a lower water ratio would cause the ink unstable. In addition, the p-type ink should be deposited underneath the n-type film as the n-type film would be destroyed by the p-type ink if they were deposited in the reverse order. As shown in Fig. 22, the device produces PCE of 9.0% with V_oc of 0.57 V, J_sc of 26.22 mA/cm², and FF of 0.61. This work demonstrates that high-quality p-type CQD films are able to be fabricated by blade coating, and the preparation of better p-type CQD inks is the prerequisite to realize higher-performance devices.

Figure 22.  Performance of solar cells in Ref. [68]: (a) current density versus voltage curves before (red) and after (blue) light soaking and (b) EQE and integrated J_sc curves. Reproduced with permission [68].

The best-performed blade-coated CQD solar cell is shown in Fig. 23 (a), which was demonstrated by Kirmani et al. in 2018 [69]. They also devised that a simple post-deposition treatment, termed oxygen-doping, could enable ambient manufacturing irrespective of seasonal humidity variations. The CQD ink was produced by SPLE with a mixture of PbI₂, PbBr₂, and ammonium acetate ligands [54], and the CQD solar cell was fabricated with a single blade-coating step. It not only achieved similar PCE of ~10% (Fig. 23 (b)) as that of spin-coated devices, but also resulted in a large reduction in material consumption. Taking advantage of this high-throughput coating technique, they also fabricated flexible CQD solar cells on the PET substrate and achieved the highest PCE of 8.0%.

Figure 23.  Device fabrication process and performance of the PbS CQD solar cell in Ref. [69]: (a) schematic of the blade-coating process and (b) current density versus voltage curves for solar cells of different active areas. Reproduced with permission [69].

As stated above, blade-coated CQD films from CQD inks could realize devices with comparable performance to these spin-coated ones. While superior to spin-coated films, the thickness of blade-coated films can be tuned in a broad range more easily. Especially with the development of better passivation strategies, the advantage of blade-coating methods will be more evident.

As diagrammatized in Fig. 24, dip coating is a simple and effective coating technique commonly used in various industry fields and has become a main coating method in the fabrication of thin films with a purpose-built dip coater. Under optimized conditions, dip coating not only could successfully manufacture highly uniform films, but also is advantageous to other coating techniques with regard to its simplicity in design which contributes to low costs of installation and maintenance. However, due to its difficulty in simultaneous control of solution spreading and evaporation, the viscous flow dynamics and evaporation should be carefully addressed. To do so, coating parameters and other factors, such as the withdrawal speed, temperature, airflow, and cleanliness, must be closely monitored during the coating process, which requires much deep understanding and skilled coating experience. As a matter of fact, for the CQD solar cell field, dip coating is less used in practice. It was only developed for the LbL process [41,87-97], since the self-assembly monolayer with dip coating is relatively easy. Here we will briefly review several dip-coating examples used in the CQD field.

Figure 24.  Schematic of dip coating.

Most earlier studies are focused on dye-sensitized solar cells (DSSCs). The reason is that porous TiO₂ layers are employed in DSSCs, allowing that the pores can be easily filled by dip coating. For example, Liu and Wang demonstrated that PbS CQDs could be successfully deposited onto porous TiO₂ layers by dipping the TiO₂-coated substrate into Pb⁺ and S^2– ion solutions alternatively, leading to in-place chemical reactions [90]. This is commonly termed the dip-SILAR method, where SILAR means the successive ionic layer adsorption and reaction. With a similar dip-SILAR method, Ratanatawanate et al. fabricated PbS CQDs doped TiO₂ nanotubes by dipping a TiO₂-nanotubes-coated substrate into the Pb(NO₃)₂ aqueous solution and the Na₂S aqueous solution, respectively [87]. Dip coating was later employed to fabricate CQD solar cells with the LbL protocol, where carboxylic-capped CQDs were dip-coated onto the substrate, and SSE and rinsing were conducted as well. For example, Luther et al. reported a Schottky solar cell based on PbSe CQD thin films in 2008 [41]. The PbSe CQD thin films were fabricated as follows: ITO substrates were dipped by hand into a 20 mL beaker filled with the oleate-capped PbSe CQD solution in hexane (6 mg/mL), followed by a second beaker containing 0.01 M EDT in acetonitrile. The 25 to 40 dip-coating cycles were performed to fabricate films of different thicknesses. Based on the deposited PbSe CQD films, the corresponding device obtained PCE of 2.1%. Later in 2010, Ju et al. adopted the same way to fabricate the PbS CQD active layer and thus reported a PbS CQD solar cell, which increased PCE to >3% by introducing another TiO₂ layer into the device [89]. However, there has been no report on dip coating with CQD inks. Because the thickness control is a challenge for the solvent used for the CQD ink. Solvent engineering might be conducive to a breakthrough in this realm.

Slot-die coating is another used deposition method in industry, which can be easily integrated with roll-to-roll processes, as shown in Fig. 25. However, very few studies on slot-die-coated CQD solar cells have been reported. As a comparison, slot-die-coated perovskite [98-107] and organic [108-114] solar cells have seen rapid growth. Here we will mainly introduce the general principles of slot-die coating and several application examples taken from other types of solar cells.

Figure 25.  Schematic of slot-die coating.

Solvent engineering is of special importance for slot-die coating because solvents could affect the substrate wettability, meniscus dynamics, and film drying and crystallization. For example, Yang et al. employed slot-die coating and successfully fabricated efficient and stable large-area formamidinium-cesium perovskite parallel solar modules [107]. In this work, diphenyl sulfoxide (DPSO) was added into the FA_0.83Cs_0.17PbI_2.83Br_0.17 precursor solution with a combined solvent of DMF and N-methyl-2-pyrrolidinone (NMP). They found that the DPSO additive could enlarge the nucleation energy barrier, effectively retard the natural nucleation of perovskite during the coating process, and stabilize the wet precursor film. The results indicate that the solvent plays a decisive role in the microscale processes. It not only affects the dynamics within the ink but also modulates the macroscale morphologies of the film. Currently, similar solvents for perovskite inks have been used in CQD inks. For example, DMF is commonly used in both CQD [59,60] and perovskite [107] inks. Other solvents or solvent mixtures, such as γ-butyrolactone (GBL) and 2-methoxy-ethanol (2-ME):DMSO, are also potential alternatives for CQD inks.

With the rising concern about the environmental friendliness of technologies, heavy metals used in solar cells, such as Pb and Cd, are facing regulations and bans in various industrial fields including solar plants, the rooftop of houses, and solar window installations of buildings. As a consequence, solar cell technologies based on heavy-metal-free materials are being developed. Zhou et al. reviewed the recent development of heavy-metal-free CQD solar cells [115], including binary and ternary silver-based CQDs (Ag₂S, Ag₂Se, AgInSe₂, and AgBiS₂) as well as ternary and quaternary copper-based CQDs (CuInS₂, CuInSe₂, ZnCuInSe, and CuInGaSe). Here we will review the main studies on the large-area fabrication of those types of solar cells.

In 2013, Panthani et al. from The University of Texas at Austin employed spray coating and realized CuInSe₂ CQD solar cells [62]. As shown in Fig. 26 (a), the fabricated device has a structure of glass/Au/CuInSe CQDs/CdS/ZnO/ITO. It shows very high V_oc of 0.85 V (Figs. 26 (b) and (c)), resulting in a small V_oc deficit of 0.61 V for a CQD bandgap of 1.46 eV. With a low-concentration CQD ink (20 mg/mL), spray coating enables a thick film (200 nm), which is difficult for other deposition techniques. However, due to the low J_sc and FF, only 0.3%−1.2% of overall PCE could be achieved. With the fast development in materials and device structures, the best CuInSe₂ CQD solar cells (in a dye-sensitized solar cell architecture) can now achieve PCE of larger than 10% [115]. It is expected that higher-performance solar cells based on the scalable fabrication method could be realized by combining those latest advances.

Figure 26.  Device structure and performance of spray-coated CuInSe₂ CQD solar cells: (a) device structure, (b) current density versus voltage curves, and (c) V_oc, J_sc, FF, and PCE values as a function of CQD bandgap. Reproduced with permission [62].

In 2018, Shen et al. developed a kind of solar paint with the composition of TiO₂ particles and ZnCuInSe CQDs [116]. Fig. 27 depicts its preparing procedure. The solar paint was screen-printed onto the fluorine-doped tin oxide (FTO) glass substrate and a dye-sensitized type of CQD solar cells was made. By optimizing the CQD load, the highest PCE of 4.2% was obtained with V_oc of 0.59 V, J_sc of 11.6 mA/cm², and FF of 0.63, respectively. Although this performance is far lower than the best-performed ZnCuInSe CQD solar cell (PCE of 13.84%) [117], it demonstrates that heavy-metal-free CQD solar cells are able to be fabricated with scalable fabrication processes. However, limited by few investigations, the potential is yet to be demonstrated and more efforts are needing to be made.

Figure 27.  Schematic procedure for the preparation of the CQD-TiO₂ solar paint. Reproduced with permission [116].

This paper reviewed recent studies on the fabrication of large-area CQD solar cells with various roll-to-roll compatible techniques. Amongst, spray coating and blade coating are shown to be promising for producing high-performance CQD solar cells. While other coating techniques, including slot-die coating and dip coating, which have been used for perovskite and organic solar cells, do not attracted much attention in the field of CQD solar cells. The performance of CQD solar cells based on those methods is still low compared with that of other solution-processed technologies, such as perovskite and organic solar cells. There is still a long way to go. However, CQDs are also promising for solar cell applications. There are two main merits which make CQD solar cells highly attractive. First, CQDs, especially lead chalcogenide CQDs, could harvest NIR radiation. By stacking CQD solar cells with Si, perovskite, and organic solar cells to form tandem solar cells, the device could collect more sunlight and achieve higher PCE. Second, CQD solar cells show excellent long-term stability. It has been verified that unencapsulated PbS CQD solar cells are stable under both shelf storage [85] and continuous operation [118]. For example, Choi et al. reported that the unencapsulated PbS CQD solar cell passivated by PbX could retain >80% of its initial efficiency after 300 h continuous operation in air [118].

However, CQD solar cells are still needing to be further investigated. Amongst, breakthroughs in fundamental understanding of the limiting factors, such as the large V_oc deficit, charge carrier dynamics, and electronic structures, are especially vital. The knowledge and experience could benefit to obtain better CQD inks as well as better device structures. Meanwhile, special attention should be paid to solvent engineering and parameter control during large-area deposition. Extensive efforts should be concentrated on the mechanism study, such as the solvent-solvent interaction, solvent-CQD interaction, and solvent-substrate interaction. And for deposition parameter control, a huge parameter space needs to be optimized. These mean that massive experimental trials are necessary, which is time and resource consuming. Here we would like to bring the readers’ attention to alternative ways of conducting scientific research by incorporating advanced information technologies, such as artificial intelligence (AI).

AI is revolutionizing every corner of our world, and Intelligent CQD Era is coming soon. The recent advances in machine learning (ML) for materials discovery, such as Materials Genome Initiative, provide new momentum for materials science [119]. ML is able to screen millions of possible candidates computationally or vast parameter spaces for scientific experiments, and predict the properties of a wide range of materials, even though the underlying theories or mechanisms behind are barely understood. On the basis of known facts, ML could further provide new insights. ML uses computer algorithms to construct mathematical models. These models can perform certain tasks, such as prediction and clustering, directly from existing data (such as images, graphs, or tables), with no need to know established physical laws.

There are already several groups on the way of empowering CQD research with AI. For example, Voznyy et al. from University of Toronto used ML to screen the reaction temperature and precursor concentration of the hot-injection synthesis of PbS CQDs and got the optimized synthesis condition for optimum monodispersity (narrow size distribution) [120]. However, it is still time consuming and resource intensive, since a large amount of data collected from day-to-day experiments are needed. Vikram et al. from University of Illinois at Urbana-Champaign proposed a more efficient predictive synthesis approach, which employed an autonomous flow reactor (as shown in Fig. 28) combining in situ spectroscopy monitoring with an artificial neural network algorithm [121]. Their setup is essentially self-driven, and even without previous knowledge of synthesis, the setup could achieve target CQDs (indium phosphide) with less than 50 experimental trials within 2 days. For more comprehensive understanding, the readers can refer to the relevant reviews [119,122]. Those experiments are mostly focused on the synthesis of CQD, while regarding to the CQD solar cell, it is much more challenging to accomplish similar self-driven setups. Cao et al. from University of Alberta attempted to use ML to analyze and predict the effects of the ink composition and film thickness on the performance of organic solar cells [123,124]. Their study showed that with AI, a comprehensive view in all directions of the parameter space (donor concentration and film thickness) could be reviewed, revealing the power of AI for solar cell research.

Figure 28.  Schematic of the autonomous flow reactor platform, including key stages of precursor mixing, nucleation reactor, growth reactor, automated sampling, inline spectroscopy, and artificial neural network module.

From the above AI applications in the materials study, we can conclude that with these joint forces, CQD solar cell research could be accelerated and new insights could be revealed. Generally, large-area coating techniques could be seamlessly integrated with AI, so that, the coating processes could be automated, optimized, and augmented with AI. We believe that the AI-powered one-step solution for CQD solar cell research will be feasible in the coming years.

This work was supported by the National Natural Science Foundation of China under Grants No. 11774304, No. 61905206, No. 12064048, and No. 11804294.

The authors declare no conflicts of interest.